Skip to content

Latest commit

 

History

History
76 lines (53 loc) · 2.31 KB

README.md

File metadata and controls

76 lines (53 loc) · 2.31 KB
Raw

TimeSeriesForecastingModelsComparison

Overview

This repository contains a comprehensive analysis and modeling of daily visitors to four different restaurants, utilizing various statistical and machine-learning techniques. The dataset, RestaurantVisitors.csv, tracks visitor counts and holiday indicators over time.

Dataset Description

Objectives

The primary goal of this project is to predict daily restaurant visitors using various modeling techniques, comparing their performance based on the Root Mean Squared Error (RMSE) as the primary metric.

Methodology

Modeling Techniques

Statistical Models

  1. Random Walk
  2. Random Walk with Drift
  3. ARIMA (AutoRegressive Integrated Moving Average)
  4. SARIMA (Seasonal ARIMA)
  5. AR (AutoRegressive)
  6. MA (Moving Average)
  7. ARIMAX (ARIMA with exogenous variables)
  8. SARIMAX (Seasonal ARIMA with exogenous variables)
  9. Exponential Smoothing
    • Simple Exponential Smoothing
    • Holt’s Linear Trend Model
    • Holt-Winters Seasonal Model
  10. VAR (Vector AutoRegression)
  11. VARMA (Vector AutoRegressive Moving Average)
  12. VARMAX (Vector AutoRegressive Moving Average with exogenous variables)
  13. ETS (Error, Trend, Seasonality) Models

Classical Machine Learning Models

  1. Linear Regression
  2. Support Vector Regression (SVR)
  3. Decision Trees Regressor
  4. Random Forest
  5. Gradient Boosting
  6. XGBoost (Extreme Gradient Boosting)
  7. LightGBM (Light Gradient Boosting Machine)
  8. CatBoost
  9. K-Nearest Neighbors (KNN)

Deep Learning Models

  1. Vanilla RNN(Recurrent Neural Network)
  2. LSTM (Long Short-Term Memory)
  3. GRU (Gated Recurrent Unit)
  4. LSTM-CNN (CNN combined with LSTM)
  5. Temporal Convolutional Networks (TCN)
  6. Transformer Models
    • Transformers (e.g., BERT, GPT)
    • Time Series Transformers
  7. Seq2Seq Models

Specialized Models

  1. FBProphet (Facebook's forecasting tool)
  2. DeepAR (Deep Autoregressive Models for Forecasting)
  3. Temporal Fusion Transformers (TFT)
  4. Matrix Factorization Models (for multivariate time series)

Performance Evaluation

The models were evaluated based on the RMSE as the primary metric to identify the best-performing model for predicting restaurant visitors.

Result